AI Coding Agent reports changes applied, but live preview iframe remains stuck on stale build (Cache / Cloud Run sync issue)

Issue Summary:
In my Google AI Studio Build project, the AI Coding Agent repeatedly reports that requested code modifications (removing a specific asset/card from the UI and backend) have been executed, built, and verified successfully. However, the interactive application in the preview window (and the shared preview URL) continues to render the outdated version without reflecting the changes.

Observed Behavior:

  1. The user requested to remove an unwanted token (ZEC) and replace it with top-volume assets (JUP and USDC).
  2. The agent reported completing the task, compiling without errors (npm run build, tsc --noEmit), and restarting the dev server.
  3. Even after hard-refreshing the browser (Ctrl + F5 / Cmd + Shift + R), clearing local storage, and restarting the browser, the preview iframe still displays the old 6th asset card (ZEC) with previous baseline values.
  4. Checking the live endpoints via curl confirms a discrepancy between the local container state and what the external Cloud Run reverse proxy / CDN is serving to the client browser.

Requested Assistance:
Could the engineering or support team check if there is an issue with the Cloud Run container synchronization, stale static asset caching in the proxy layer, or a frozen container instance preventing dev server updates from reaching the client iframe?

Thank you.